The Large APEX Bolometer Camera (LABOCA) is a new 295-bolometer submillimeter camera operating at 870 μm that has been commissioned for use on the 12 m Atacama Pathfinder Experiment (APEX) telescope. LABOCA offers unprecedented mapping capabilities for submillimeter continuum emission due to APEX's efficiency and the excellent atmospheric transmission at the site. LABOCA is designed to map large areas of the sky rapidly without the need for a chopping secondary mirror by using fast scanning techniques and correlating signals across the array to remove atmospheric contributions.

1. Astronomy & Astrophysics manuscript no. 1454light c ESO 2009
March 7, 2009
The Large APEX Bolometer Camera LABOCA
G. Siringo1 , E. Kreysa1 ,
A. Kovács1 , F. Schuller1 , A. Weiß1 ,
W. Esch 1 , H.-P. Gemünd1 , N. Jethava2 , G. Lundershausen1 , A. Colin3 ,
R. Güsten1 , K. M. Menten1 , A. Beelen4 , F. Bertoldi5 , J. W. Beeman6 , E. E. Haller6 ,
arXiv:0903.1354v1 [astro-ph.IM] 7 Mar 2009
1 Max-Planck-Institut für Radioastronomie, Auf dem Hügel 69, 53121 Bonn, Germany
2 National Institute of Standards and Technology, Boulder, CO 80305, USA
3 Instituto de Fisica de Cantabria (CSIC-UC), Avda. Los Castros, 39005 Santander, Spain
4 Institut d’Astrophysique Spatiale, bât 121 - Université Paris-Sud, 91405 Orsay Cedex, France
5 Argelander-Institut für Astronomie, University of Bonn, Auf dem Hügel 71, 53121 Bonn, Germany
6 Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
Received 2 December 2008 / Accepted 21 January 2009
ABSTRACT
The Large APEX Bolometer Camera, LABOCA, has been commissioned for operation as a new facility instrument at the
Atacama Pathfinder Experiment 12 m submillimeter telescope. This new 295-bolometer total power camera, operating in the
870 µm atmospheric window, combined with the high efficiency of APEX and the excellent atmospheric transmission at the site,
offers unprecedented capability in mapping submillimeter continuum emission for a wide range of astronomical purposes.
Key words. Instrumentation: detectors – Instrumentation: photometers – Techniques: bolometers – Techniques: array – Techniques:
submillimeter
1. Introduction given dust-to-gas relation) in the telescope beam. Accordingly,
the (sub)millimeter flux can be converted into dust/gas masses,
1.1. Astronomical motivation when the temperature T is assumed or constrained via additional
Millimeter and submillimeter wavelength continuum emission is far-infrared measurements.
a powerful probe of the warm and cool dust in the Universe. For Such mass determination is one of the core issues of
temperatures below ∼40 K, the peak of the thermal continuum (sub)millimeter photometry. We would like to illustrate its
emission is at wavelengths longer than 100 µm (or at frequencies paramount importance with a few examples: (sub)millimeter
lower than 3 THz), i.e. in the far-infrared and (sub)millimeter1 wavelengths mapping of low-mass star-forming regions in
range. A number of atmospheric windows between 200 GHz and molecular clouds have determined the dense core mass spec-
1 THz make ground-based observations possible over a large trum, (in nearby regions) down to sub-stellar masses, and inves-
part of this range from high-altitude, dry sites. tigated its relationship to the Initial Mass Function (Motte et al.
Specifically, thermal dust emission is well described by a 1998). These studies can be extended to high-mass star-forming
gray body spectrum, with the measured flux density S ν at fre- regions (e.g. Motte et al. 2007; Johnstone et al. 2006). Because
quency ν expressed as: of the larger distances to rarer, high-mass embedded objects,
2h ν3 even relatively shallow surveys are capable of detecting pre-
S ν = ΩS ′ (1 − e−τ) stellar cluster clumps with a few hundreds of solar masses of
c2 ehν/kT − 1 material at distances as far as the Galactic center. Masses and
where h and k are Planck’s and Boltzmann’s constants respec- observed sizes yield radial density distribution profiles for pro-
tively, c is the speed of light, ΩS ′ is the apparent source solid tostellar cores that can be compared with theoretical models
angle (the size of the physical source convolved with the tele- (Beuther et al. 2002).
scope beam) and τ is the optical depth, which varies with fre- (Sub)millimeter continuum emission is also a remarkable
quency. In the (sub)millimeter range, the emission is almost al- tool for the study of the distant Universe. The thermal dust
ways optically thin with τ ∝ Nνβ , where β is in the range 1– emission in active galaxies is typically fueled by short-lived,
2 typically (see, e.g., the Appendix of Mezger et al. 1990 or high-mass stars, therefore the far-infrared luminosity provides
Beuther et al. 2002 for a more thorough discussion). Here, N is a snapshot of the current level of star-formation activity. Deep
the number of dust particles (or number of nucleons assuming a (sub)millimeter observations in the Hubble Deep Field, with
Send offprint requests to: G. Siringo the Submillimeter Common User Bolometer Array (SCUBA,
e-mail: gsiringo@mpifr-bonn.mpg.de Holland et al. 1999) on the James Clerk Maxwell Telescope,
1 Hereafter we will use the term (sub)millimeter when referring, in attracted considerable attention with the detection of a few
general, to the millimeter and submillimeter wavelengths range; we will sources without optical or near-infrared counterpart. Sensitive
use submillimeter when strictly referring to wavelengths from one mil- measurements found dust in high redshift sources and even in
limeter down to the far-infrared. some of the farthest known objects in the Universe (see, e.g.,

2. 2 G. Siringo et al.: The Large APEX Bolometer Camera LABOCA
the Atacama Pathfinder Experiment 12 m telescope, APEX2 . It
is the most ambitious camera in a long line of developments of
the MPIfR bolometer group, which has delivered instruments
of increasing complexity to the IRAM 30 m telescope (Kreysa
1990): single beam receivers were supplanted by a 7-element
system (Kreysa et al. 1999), which eventually gave way to the
MAx-Planck Millimeter BOlometer (MAMBO) array, whose
initial 37 beams have grown to 117 in the latest incarnation
(Kreysa et al. 2002). The group also built the 37-element 1.2 mm
SEST Imaging Bolometer Array (SIMBA) for the Swedish/ESO
Submillimeter telescope (Nyman et al. 2001) and the 19-element
870 µm for the 10 m Heinrich-Hertz Telescope (a.k.a. SMTO,
Martin & Baars 1990).
The main obstacle to observations at these wavelengths is
posed by Earth’s atmosphere, which is seen as a bright emit-
ting screen by a continuum total power detector, as LABOCA’s
bolometers are. This is largely due to the emission of the water
vapor present in the atmosphere with only small contributions
from other components, like ozone. Besides, the atmosphere is a
turbulent thermodynamic system and the amount of water vapor
along the line of sight can change quickly, giving rise to instabili-
ties of emission and transmission, called sky noise. Observations
Fig. 1. Wiring side of a naked LABOCA array. Each light-green from ground based telescopes have to go through that screen,
square is a bolometer. therefore requiring techniques to minimize those effects.
The technique most widely used is to operate a switching de-
vice, usually a chopping secondary mirror (hereafter called wob-
bler) to observe alternatively the source and a blank sky area
Carilli et al. 2001; Bertoldi et al. 2003; Wang et al. 2008), re- close to it, at a frequency higher than the variability scale of
vealing star-formation rates (SFRs) that are hundreds of times the sky noise. This method, originally introduced for observa-
higher than in the Milky Way today. These detections are possi- tions with single pixel detectors, is today also used with arrays
ble because of the so-called negative K-correction first discussed of bolometers. Although it can be very efficient to reduce the
by Blain & Longair (1993): the warm dust in galaxies is typ- atmospheric disturbances during observations, it presents some
ically characterized by temperatures around 30–60 K. Its ther- disadvantages: among others, the wobbler is usually slow (1 or
mal emission dominates the spectral energy distribution (SED) 2 Hz) posing a strong limitation to the possible scanning speed.
of luminous and ultra-luminous infrared galaxies (LIRGs and
LABOCA has been specifically designed to work without
ULIRGs), which have maxima between 3 and 6 THz as a result.
a wobbler and using a different technique, which works par-
Because of the expansion of the Universe, the peak of the emis-
ticularly well when using an array of detectors, to remove the
sion shifts toward the lower frequencies with increasing cos-
atmospheric contribution. This technique, called fast scanning
mological distance, thus counteracting the dimming and ben-
(Reichertz et al. 2001), is based on the idea that, when observing
efitting detection at (sub)millimeter wavelength. Consequently,
with an array, the bolometers composing the array look simulta-
flux-limited surveys near millimeter wavelengths yield flat or
neously at different points in the sky, therefore chopping is no
nearly flat luminosity selection over much of the volume of
more needed. A modulation of the signal, still required to iden-
the Universe (Blain et al. 2002). As such, (sub)millimeter wave-
tify the astronomical source through the atmospheric emission,
lengths allow unbiased studies of the luminosity evolution and,
is produced by scanning with the telescope across the source.
therefore, of the star-formation history of galaxies over cosmo-
The atmospheric contribution (as well as part of the instrumen-
logical time-scales.
tal noise) will be strongly correlated in all bolometers and a
post-detection analysis of the correlation across the array will
1.2. Bolometer arrays for (sub)millimeter astronomy make it possible to extract the signals of astronomical interest
from the atmospheric foregrounds. Moreover, the post-detection
The (sub)millimeter dust emission is unfortunately intrinsically bandwidth depends on the scanning speed, therefore relatively
weak and the wish to measure it pushed the development of high scanning speeds are ideal (see also Kovács 2008).
detectors with the best possible sensitivity, namely bolometers The APEX telescope (Güsten et al. 2006), as the name
(Low 1961; Mather 1984). Moreover, the desirability of map- implies, serves as a pathfinder for the future large-scale
ping large areas of the sky, motivated the development of de- (sub)millimeter wavelength and (far)infrared missions, namely
tector arrays. Consequently, the last 10 years have seen an in- the Atacama Large Millimeter Array (ALMA), the Herschel
creasing effort in the development of bolometer arrays. In such Space Observatory and the Stratospheric Observatory for
instruments, a number of composite bolometers work side by Infrared Astronomy (SOFIA). Its pathfinder character is on the
side in the focal plane, offering simultaneous multi-beam cover- one hand defined by exploring wavelength windows that have
age. Since the arrival of the first arrays, developed in the early been poorly studied before, with acceptable atmospheric trans-
’90s and consisting of just 7 elements, we are witnessing a rapid
maturing of technology, reaching hundreds to a few thousand el- 2 APEX is a collaborative effort between the Max-Planck-
ements today, and the prospect of even larger bolometer arrays Institut für Radioastronomie of Bonn (MPIfR, 50%), the European
in the future. The Large APEX Bolometer Camera, LABOCA, Southern Observatory (ESO, 27%) and the Onsala Space Observatory
described in this article, is a new bolometric receiver array for (OSO, 23%)

3. G. Siringo et al.: The Large APEX Bolometer Camera LABOCA 3
/'

4. )
$
$
. 5

5. +$
+2 (
+ ,
( '6
'%/ -
/
)

6. #$
(
(
(
(
, #
(

7. ! !
# )*
)
/
/ #
1 (

8. 0
1
*
( , 0

9. #

10. 1 (/ '((
)*
3
# $ (
4
*
1 %
%
+ *
Fig. 2. Scheme of the infrastructure of LABOCA. The instrument is located in the Cassegrain cabin (dashed red line) of APEX,
remote operation is a requirement and all the communication goes through the local area network (magenta arrows) and a direct
Ethernet link between backend and bridge computers. The black lines show the flow of the bolometer signals, the orange ones show
the configuration and monitoring communication.
mission at the 5100 m altitude site. On the other hand, and more The bolometer array is mounted in a cryostat, which uses
importantly, it can perform large area mapping to identify in- liquid nitrogen and liquid helium on a closed cycle double-stage
teresting sources for ALMA follow-up studies at higher angular sorption cooler to reach an operation temperature of ∼285 mK.
resolution. Moreover, APEX produces images of both contin- The cryogenic system is discussed in §3.
uum (with LABOCA) and line emission with angular resolutions A set of cold filters, mounted on the liquid nitrogen and liq-
that neither Herschel’s nor SOFIA’s smaller telescopes (with di- uid helium shields, define the spectral passband, centered at a
ameters of 3.5 and 2.5 m respectively) can match. This provides wavelength of 870 µm and about 150 µm wide (see Fig. 5). A
a critical advantage to APEX for imaging dust and line emission monolithic array of conical horn antennas, placed in front of the
at high frequencies. bolometer wafer, concentrates the radiation onto the individual
bolometers. The filters and the horn array are presented in §4.
The cryostat is located in the Cassegrain cabin of the APEX
1.3. Instrument overview telescope (see Fig. 4) and the optical coupling to the telescope is
provided by an optical system made of a series of metal mirrors
LABOCA is an array of bolometers, operated in total-power and a lens which forms the cryostat window. The complex optics
mode, specifically designed for fast mapping of large areas of layout, manufacture and installation at the telescope in described
sky at moderate resolution and with high sensitivity and was in §2.
commissioned in May 2007 as facility instrument on APEX. It The bolometer signals are routed through low noise, unity
is a very complex system, composed of parts that originate in gain Junction Field Effect Transistor (JFET) amplifiers and to
a variety of fields of technology, in particular optics, high vac- the outside of the cryostat along flexible flat cables. Upon exit-
uum, low temperature cryogenics, digital electronics, computer ing the cryostat, the signals pass to room temperature low noise
hardware and software, and others. A general view of the in- amplifiers and electronics also providing the AC current for bias-
frastructure is shown in the block diagram of Fig. 2. The heart ing the bolometers and performing real time demodulation of the
of LABOCA is its detector array made of 295 semiconducting signals. The signals are then digitized over 16 bits by 4 data ac-
composite bolometers (see Fig. 1, Fig. 6). A description of the quisition boards providing 80 analog inputs each, mounted in the
detector array design and manufacture is provided in §5. backend computer. The backend software provides an interface

11. 4 G. Siringo et al.: The Large APEX Bolometer Camera LABOCA
Fig. 3. Overview of the optics. On the left, the APEX telescope. On the right, a zoom on the tertiary optics installed in the Cassegrain
cabin. See also Fig. 4.
to the telescope’s control software, used to set up the hardware, 2. Tertiary optics
and a data server for the data output. The acquired data are then
digitally filtered and downsampled to a lower rate in real time 2.1. Design and optimization
by the bridge computer and finally stored in MB-FITS format The very restricted space in the Cassegrain cabin of APEX and a
(Muders et al. 2006) by the FITS writer embedded in the tele- common first mirror (M3) with the APEX SZ Camera (ASZCa,
scope’s control software. Another computer, the frontend com- Schwan et al. 2003) introduced many boundary conditions into
puter, is devoted to monitor and control most of the electronics the optical design. Eventually, with the help of the ZEMAX3 op-
embedded into the receiver (e.g. monitoring of all the tempera- tical design program, a satisfactory solution was found, featuring
ture stages, controlling of the sorption cooler, calibration unit) three aspherical off-axis mirrors (M3, M5, M7), two plane mir-
and provides an interface to the APEX control software, allow- rors (M4, M6) and an aspherical lens acting as the entrance win-
ing remote operation of the system. Discussions of the cold and dow of the cryostat (see Fig. 3). Meeting the spatial constraints,
warm readout electronics, plus the signal processing, are found without sacrificing optical quality, is facilitated considerably by
in §6 and §7, respectively. the addition of plane mirrors. The design of the optics was made
at the MPIfR in coordination with N. Halverson4 with respect to
In §8 we describe the sophisticated observing techniques sharing the large M3 mirror with the ASZCa experiment.
used with LABOCA, some of them newly developed. The in- The maximum possible field diameter of APEX, as limited
strument performance on the sky is described in §9, along with by the diameter of the Cassegrain hole of the telescope’s pri-
information on sensitivity, beam shape and noise behavior. mary, is about 0.5 degrees. LABOCA, with its 295 close-packed
fully efficient horns, covers an almost circular field of view
The reduction of the data is performed using a new data re- (hereafter FoV) of about 0.2 degrees in diameter (or about 100
duction software included in the delivery of LABOCA as facility square arcminutes). The task of the tertiary optics is to trans-
instrument, the BoA (Bolometer array data Analysis) data reduc- form the f-ratio from f/8 at the Cassegrain focus to f/1.5 at the
tion software package. An account of on-line and off-line data horn array, while correcting the aberrations over the whole FoV
reduction is given in §10. of LABOCA under the constraint of parallel output beams. The
Some of the exciting science results already obtained with 3 http://www.zemax.com/
LABOCA or expected in the near future are outlined in §11. 4 Formerly at University of California at Berkeley, now at University
Our plans for LABOCA’s future are briefly presented in §12. of Colorado, Boulder, USA

12. G. Siringo et al.: The Large APEX Bolometer Camera LABOCA 5
and the ASZCa experiment. M3 has a diameter of 1.6 m and a
surface accuracy of 18 µm rms but LABOCA uses only the in-
ner 80 cm disk. It is attached to a bearing on the floor of the
Cassegrain cabin, aligned to the optical axis of APEX. Operators
can manually rotate the mirror in order to direct the telescope
beam to LABOCA or to ASZCa alternatively (see Fig. 4).
Mirrors M4 and M6 are flat, have a diameter of 42 cm
and 26 cm, respectively (manufactured by Kugler5, Salem,
Germany). They are of optical quality and are both affixed to the
ceiling of the cabin. In a near future, mirror M6 will be replaced
by the reflection-type half-wave plate of the PolKa polarimeter
(Siringo et al. 2004).
Mirrors M5 and M7 are off-axis aspherics, both 50 cm in di-
ameter, and are affixed to the floor of the cabin. They have been
designed and manufactured at the MPIfR and have a surface ac-
curacy of 7 and 5 µm rms respectively.
2.3. Installation and alignment
All the mirrors of LABOCA (with exception of M3) and the
receiver itself are mounted on hexapod positioners (see Fig. 4)
provided by VERTEX Antennentechnik6 (Duisburg, Germany).
Each hexapod is made of an octahedral assembly of struts and
has six degrees of freedom (x, y, z, pitch, roll and yaw). The
lengths of the six independent legs can be changed to posi-
tion and orient the platform on which the mirror is mounted.
VERTEX provided a software for calculating the required leg
extensions for a given position and orientation of the platforms.
A first geometrical alignment was performed during the first
week of September 2006, using a double-beam laser on the opti-
cal axis of the telescope and plane replacement mirrors in place
of the two active mirrors M5 and M7. The alignment has been
checked using the bolometers and hot targets (made of absorb-
ing7 material) at different places along the beam, starting at the
focal plane and following the path through all the reflections up
to the receiver’s window. The alignment has been furthermore
verified and improved in February 2008.
3. Cryogenics
3.1. Cryostat
The bolometer array of LABOCA is designed to be operated at
a temperature lower than 300 mK. This temperature is provided
by a cryogenic system made of a wet cryostat, using liquid nitro-
gen and liquid helium, in combination with a two-stage sorption
Fig. 4. Top: The mirrors affixed to the floor of the Cassegrain cooler. A commercial 8-inch cryostat, built by Infrared Labs8
cabin of APEX. Left to right: M7, M5, M3 and one of the mir- (Tucson, AZ, USA), has been customized at the MPIfR to ac-
rors of the ASZCa experiment. In this picture, mirror M3 is posi- commodate the double-stage sorption cooler, the bolometer ar-
tioned for ASZCa. Bottom: The receiver, M3 in the position for ray, cold optics and cold electronics. A high vacuum in the cryo-
LABOCA, M5, M6 and M7. See also Fig. 3. stat is provided by an integrated turbomolecular pump backed by
a diaphragm pump. Operational vacuum is reached in one single
day of pumping.
final design is diffraction limited even for 350 µm wavelength, The cryostat incorporates a 3-liter reservoir of liquid nitro-
the Strehl ratio is better than 0.994 and the maximum distortion gen and a 5-liter reservoir of liquid helium. After producing
at the focal plane is less than 10% over the entire FoV (see also high-vacuum (∼10−6 mbar), the cryostat is filled with the liquid
Fig. 10). cryogens. The liquid nitrogen is used to provide thermal shield-
ing at 77 K in our labs in Bonn (standard air pressure, 1013 mbar)
5 http://www.kugler-precision.com/
2.2. Manufacture
6 http://www.vertexant.de/
Mirror M3, an off-axis paraboloid, has been manufactured by 7 ECCOSORB AN, Emerson Cuming, Rundolph, MA, USA,
a machine shop at the Lawrence Berkeley National Laboratory http://www.eccosorb.com
(LBNL, Berkeley, CA, USA) and is common to both LABOCA 8 http://www.irlabs.com/

13. 6 G. Siringo et al.: The Large APEX Bolometer Camera LABOCA
1.0
and at 73.5 K at APEX (5107 m above the sea level) where the air
pressure is almost one half of the standard one (about 540 mbar).
The liquid helium provides a thermal shielding at 4.2 K at 0.8
standard pressure and 3.7 K at the APEX site. To keep it oper-
ational, the system must be refilled once per day. The refilling
transmission
0.6
operation requires about 20 minutes.
0.4
3.2. Sorption cooler
The cryostat incorporates a commercial two-stage closed-cycle 0.2
sorption cooler, model SoCool (Duband et al. 2002) manufac-
372
tured by Air-Liquide9 (Sassenage, France). In this device, a 4 He 313
0.0
sorption cooler is used to liquefy 3 He gas in the adjacent, ther- 250 300 350 400 450
frequency [GHz]
mally coupled, 3 He cooler. The condensed liquid 3 He is then
sorption pumped to reach temperatures as low as 250 mK, in Fig. 5. Spectral response of LABOCA, relative to the maximum.
the absence of a thermal load. Therefore, the double stage de- The central frequency is 345 GHz, the portion with 50% or more
sign makes it possible to cool the bolometer array down to a transmission is between 313 and 372 GHz.
temperature lower than 300 mK starting from the temperature
of the liquid helium bath at atmospheric pressure. This makes
4. Cold optics
the maintenance of the system much simpler than that of other
systems, where pumping on the liquid helium bath is required. 4.1. Passband definition
The two sorption coolers are closed systems, which means they
do not require any refilling of gas and can be operated from the Inside the cryostat, a set of cold filters, mounted on the liq-
outside of the cryostat, simply by applying electrical power. uid nitrogen and liquid helium shields define the spectral pass-
band, centered at a wavelength of 870 µm (345 GHz) and about
To keep the bolometers at operation temperature, the sorp- 150 µm (60 GHz) wide (see Fig. 5). The filters have been de-
tion cooler needs to be recycled. The recycling is done by appli- signed and manufactured at MPIfR in collaboration with the
cation of a sequence of voltages to the electric lines connected to group of V. Hansen (Theoretische Elektrotechnik12, Bergische
thermal switches and heaters integrated in the sorption cooler. A Universität Wuppertal, Germany) who provided theoretical sup-
typical recycling procedure requires about two hours and can be port and electromagnetic simulations. The passband is formed
done manually or in a fully automatic way controlled by the fron- by an interference filter made of inductive and capacitive meshes
tend computer (see Sect. 7.4). At the end of the recycling pro- embedded in polypropylene. The low frequency edge of the band
cess, both gases, 4 He and 3 He, have been liquefied and the con- is defined by the cut-off of the cylindrical waveguide of each
trolled evaporation of the two liquids provides a stable temper- horn antenna (see also Sect. 4.2). A freestanding inductive mesh
ature for many hours. After the recycling of the sorption cooler, behind the window-lens provides shielding against radio inter-
the bolometer array reaches 285 mK. The hold time of the cooler, ference.
usually between 10 and 12 hours, strongly depends on the pa-
rameters used during the recycling procedure. The end temper-
ature is a function of elevation and can be affected by telescope 4.2. Horn array
movements, leading to temperature fluctuations 500 µK within
∼ A monolithic array of conical horn antennas, placed in front of
one scan, during regular observations, and 3 mK for wide ele-
∼ the bolometer wafer, concentrates the radiation onto the bolome-
vation turns (e.g. during skydips, see Sect. 8.2.3). ters. 295 conical horns have been machined into a single alu-
minum block by the MPIfR machine shop. In combination with
the tertiary optics, the horn antennas are optimized for coupling
3.3. Temperature monitor to the telescope’s main beam at a wavelength of 870 µm . The
grid constant of the hexagonal array is 4.00 mm. Each horn
The cryostat of LABOCA incorporates 8 thermometers to mea- antenna feeds into a circular wave guide with a diameter of
sure the temperature at the different stages: liquid nitrogen, liq- 0.54 mm, acting as a high-pass filter.
uid helium, the two sorption pumps, the two thermal switches,
evaporator of the 4 He and evaporator of the 3 He. Two LS21810
devices (Lake Shore Inc., Westerville, OH, USA) are used to 5. Detector
monitor the thermometers and to apply the individual tempera- 5.1. Array design and manufacture
ture calibrations in real-time. The temperature of the 3 He stage
is measured with higher accuracy with the use of a resistance The bolometer array of LABOCA is made of 295 composite
bridge AVS-4711 (Picowatt, Vantaa, Finland), with an error of bolometers arranged in an hexagonal layout consisting of a cen-
±5 µK . Control and monitor of the cryogenic equipment can be ter channel and 9 concentric hexagons (see Fig. 1 and Fig. 6).
done remotely via the frontend computer (see Sect. 7.4). The array is manufactured on a 4-inch silicon wafer coated on
both sides with a silicon-nitride film by thermal chemical va-
por deposition. On one side of the wafer, 295 squares are struc-
tured into the silicon-nitride film used as a mask for the alka-
9 http://www.airliquide.com/
10 http://www.lakeshore.com/temp/mn/218po.html
line KOH etching of the silicon, producing freestanding, un-
11 http://www.picowatt.fi/avs47/avs47.html 12 http://www.tet.uni-wuppertal.de/

14. G. Siringo et al.: The Large APEX Bolometer Camera LABOCA 7
Fig. 6. Pictures of the bolometer array of LABOCA. Left: A detail of the array mounted in its copper ring. Some bonding wires are
visible. Center: The side of the array where the bolometer cavities are etched in the silicon wafer. Right: Wiring side of the array.
The thermistors are visible as small cubes on the membranes. One broken membrane is visible on the top left corner. See also Fig. 1.
manium doped by neutron-transmutation in a nuclear reactor.
The thermistors, needed to measure the temperature of the ab-
sorber, are soldered on the bolometer membranes to gold pads
connected to the outer edge of the silicon wafer through a pair
of patterned niobium wires, superconducting at the operation
temperature (see Fig. 7 and Fig. 6, right). The soldering of the
thermistors was the only manual step in the manufacture of the
bolometers. LABOCA has so-called flatpack NTD thermistors,
which have two ion implanted and metalized contacts on one
side of the thermistor block. The chips are optimized to work at
a temperature lower than 300 mK, where they show an electric
impedance in the range of 1 to 10 MOhms.
6. Cold electronics
6.1. Bias resistors
Fig. 7. Wiring side of one bolometer of LABOCA, seen under a Given the high impedance of the NTD thermistors, the elec-
microscope. The membrane is the green colored area. The black tric scheme of the first bolometer circuitry requires very high
box is the NTD thermistor. The wiring is made of gold (yellow) impedance load resistors, which are needed to current bias the
and niobium (gray) thin metal layers. See also Fig. 6. bolometers. These bias resistors are 312 identical 30 MOhm
chips, made of a nichrome thin film deposited on silicon sub-
strate (model MSHR-4 produced by Mini-Systems Inc.13, N.
structured silicon-nitride membranes, only 400 nm thick (see the Attleboro, MA, USA) mounted on 12 identical printed circuit
center picture in Fig. 6). On the other side, the wiring is cre- boards, to form 12 groups of 26 resistors. This configuration is
ated by microlithography of niobium and gold thin metal lay- reflected in the following distribution of the bolometer signals.
ers. The bolometer array is mounted inside a gold-coated copper The circuit boards are mounted on the same copper ring which
ring and is supported by about 360 gold bonding thin wires (see holds the bolometer array (see Sect. 5.1) and electrically con-
Fig. 6, left), providing the required electrical and thermal con- nected to the bolometers through miniature RF filters.
nection. This copper ring also serves as a mount for the back-
short reflector, at λ/4 distance from the array, and 12 printed
circuit boards hosting the load resistors and the first electronic 6.2. Junction Field Effect Transistors (JFETs) source
circuitry (see Sect. 6). followers
The high impedance of the bolometers makes the system sen-
5.2. Bolometer design and manufacture sitive to microphonic noise pickup, therefore JFETs (Toshiba
2SK369) are used as source followers in order to decrease the
The description of a composite bolometer can be simplified as impedance of the electric lines down to a few kOhms before
the combination of two elements: an extremely sensitive tem- they reach the high gain amplification units at room temper-
perature sensor, called thermistor, and a radiation absorber. In ature. Following the wiring scheme of the bias resistors, the
the bolometers of LABOCA, the absorbing element is made of a JFETs are also in groups of 26 soldered onto 12 printed circuit
thin film of titanium deposited on the unstructured silicon-nitride boards, electrically connected to the corresponding bias resis-
membranes. LABOCA uses neutron-transmutation doped ger- tors by 12 flat cables made of manganin traces embedded in
manium semiconducting chips (NTD, Haller et al. 1982) as ther-
mistors. The NTD thermistors are made from ultra-pure ger- 13 http://www.mini-systemsinc.com/

15. 8 G. Siringo et al.: The Large APEX Bolometer Camera LABOCA
Kapton14 (manufactured by VAAS Leiterplattentechnologie15, and to monitor the digital reference signals (sync/blank) of the
Schwäbisch-Gmünd, Germany), thermally shunted to the liquid wobbler. For the time synchronization of the data to the APEX
helium tank. The 12 JFET boards are assembled in groups of control software (APECS, Muders et al. 2006) the data acqui-
three into four gold-coated copper boxes, thermally connected sition system is equipped with a precision time interface (PCI-
to the liquid nitrogen tank. Inside each box, during regular oper- SyncClock3220 from Brandywine Communications, Tustin, CA,
ation, the 78 JFETs are self-heated to a temperature of about USA) synchronized to the station GPS clock via IRIG-B21 time
110 K, where they show a minimum of their intrinsic noise. code signal.
Through the connections in the JFET boxes, the wiring scheme The AC bias reference frequency is provided by the data ac-
of 12 groups of 26 channels is translated to a new scheme of 4 quisition system as a submultiple of the sampling frequency, thus
groups of 80 channels. synchronizing the bias to the data sampling. Typical values used
for observations are 1 kHz for the sampling rate and 333 Hz for
the AC bias. The backend computer has two network adapters:
7. Warm electronics and signal processing one is connected to the local area network, the other one is exclu-
sively used for the output data stream and is connected in a pri-
7.1. Amplifiers
vate direct network with the bridge computer (see also Sect. 7.3).
The 312 channels exiting the cryostat of LABOCA are dis- The data acquisition software is entirely written using
tributed to 4 identical, custom made, amplification units, pro- LabVIEW22 (National Instruments). The drivers for the data ac-
viding 80 channels each. Of these, 295 are bolometers, 17 are quisition hardware are provided by the NI-DAQmx23 package.
connected to 1 MOhm resistors mounted on the bolometer ring Custom drivers for LabVIEW have been developed to access
(used for technical purposes, like noise monitoring and calibra- the GPS clock interface. The backend software runs a server to
tions) and the remaining 8 are not connected. Each amplifica- stream the output data to the bridge computer and allows remote
tion unit is made of 16 identical printed circuit boards and each control and monitoring of the operation via a CORBA24 object
board provides 5 low noise, high gain amplifiers. Each unit also interfaced to the APECS through the local area network.
includes a low noise battery used to generate the bias voltage and
the circuitry to produce the AC biasing and perform real time de-
7.3. Anti-aliasing filtering and downsampling
modulation of the 320 signals. The AC bias reference frequency
is not internally generated but is provided from the outside, thus The amplification units of LABOCA use the bias signal, which
allowing synchronization of the biasing to an external frequency is a square waveform, as reference to operate real-time de-
source. modulation of the AC-biased bolometer signals. Therefore, all
Each amplification unit is equipped with a digital interface the frequencies present in the bolometer readout lines end up
controlled by a microprocessor programmed to provide remote aliased around the odd-numbered harmonics of the bias fre-
control of the amplification gain and of the DC offset removal quency. Microphonics pickup by the high-impedance bolome-
procedure. This is required because LABOCA is a total power ters at a few resonant frequencies can produce a forest of lines
receiver and the signals carry a floating DC offset which could in the final readout, polluting even the lower part of the post-
exceed the dynamic range of the data acquisition system. To detection frequency band, where the astronomical signals are ex-
avoid saturation, therefore, the DC offsets are measured and sub- pected. To overcome this, we introduced an intermediate stage
tracted from the signals at the beginning of every integration. into the sampling scheme, the so-called bridge computer. The
The values of the 320 offsets are temporarily stored in a local bolometer signals, acquired by the backend at a relatively high
memory and, at the end of the observation, are written into the sampling rate (usually 1 kHz), well above the rolling-off of the
corresponding data file for use in the data reduction process. The anti-alias filters embedded in the amplifiers, are sent to the
digital lines use the I2 C protocol16 and are accessible remotely bridge computer where they are digitally low-pass filtered and
via the local network through I2 C-to-RS23217 interfaces con- then downsampled to a much lower rate (usually 25 Hz), more
trolled by the frontend computer (see Sect. 7.4). The amplifica- appropriate for the astronomical signals produced at the typical
tion gain can be set in the range from 270 to 17280. scanning speeds (see also Sect. 8). The digital real-time anti-
alias filtering and downsampling is performed by a non-recursive
convolution filter with a Nutall window such that its rejection
7.2. Data acquisition
at the Nyquist frequency is ∼3 dB and falls steeply to ∼100 dB
The 320 output signals from the 4 amplification units are dig- soon beyond that. The bias reference frequency was accordingly
itized over 16 bits by 4 multifunction data acquisition (DAQ) selected to maximize the astronomically useful post-detection
boards (National Instruments18 M-6225-PCI), providing 80 ana- bandwidth. The resulting bolometer signals are generally white
log inputs each and synchronized to the same sample clock by between 0.1–12.5 Hz and free of unwanted microphonic inter-
a RTSI19 bus . The maximum data sampling is 2500 Hz and the ference (see Fig. 8).
dynamic range can be selected over 5 predefined ranges. The To seamlessly integrate the bridged readout into the APEX
four boards provide also 24 digital input/output lines each, some control system, the bridge computer also acts as a fully func-
of them used for the generation of the bias reference frequency tional virtual backend that forwards all communication between
the actual backend computer and the control system, while in-
14 DuPont, http://www2.dupont.com/Kapton/en_US/
15 http://www.vaas-lt.de/ 20 http://www.brandywinecomm.com
16 Inter-Integrated Circuit, a serial
bus to connect hardware devices. 21 Inter Range Instrumentation Group, standardized time code format.
17 Recommended Standard 232, a standard for serial binary data com- 22 http://www.ni.com/labview/
munication. 23 http://www.ni.com/dataacquisition/nidaqmx.htm
18 http://www.ni.com/ 24 Common Object Request Broker Architecture, a set of standards
19 Real-Time System Integration, a bus used to share and exchange which define the protocol for interaction between the objects of a dis-
timing and control signals between multiple boards. tributed system

16. G. Siringo et al.: The Large APEX Bolometer Camera LABOCA 9
Fig. 8. Noise spectra for selected bolometers on a room temper-
ature absorber (capped optics). The signals, sampled at 1 kHz,
were downsampled in real-time to 25 Hz by the bridge computer.
The spectra are free of microphonic pick-ups and show the 1/ f
noise onset at ∼0.1 Hz.
tercepting and reinterpreting any commands of interest for the
downsampling scheme.
7.4. Frontend computer
The so-called frontend computer communicates with the hard-
ware of LABOCA through the local area network. It is devoted
to monitor and control most of the electronics of the instrument
(e.g. monitoring of all the temperature stages, control of the
sorption cooler, calibration unit, power lines. . . ) and also pro-
vides a CORBA object for interfacing to APECS, allowing re-
mote operation of the system. The frontend software is entirely
written using LabVIEW and custom drivers have been developed
for some hardware devices embedded in LABOCA.
8. Observing modes
Fig. 9. Examples of raster-spiral patterns. The two plots show
8.1. Mapping modes the scanning pattern of the central beam of the array in horizon-
In order to reach the best signal-to-noise ratio using the fast tal coordinates. The compact pattern shown in the top panel is
scanning technique (Reichertz et al. 2001) with LABOCA, the optimized to map the field of view of LABOCA. The large scale
frequencies of the signal produced by scanning across the map shown in the bottom panel consists of 25 raster positions
source need to match the white noise part of the post-detection and covers a field of 0.5 × 0.5 degrees. The complete scan re-
frequency band (0.1–12.5 Hz, see Fig. 8), mostly above the quired 19 minutes and the final map covers about 2000 × 2000
frequencies of the atmospheric fluctuations. Thus, with the arcseconds with uniform residual rms noise.
∼19′′ beam (see Sect. 9.3), the maximum practical telescope
scanning speed for LABOCA is about 4′ /s. This is also the lim-
iting value to guarantee the required accuracy in the telescope
position information of each sample. The minimum scanning and 35 s integration time, both producing fully sampled maps
speed required for a sufficient source modulation depends on the of the whole FoV with scanning velocities limited between 1′ /s
′
atmospheric stability and on the source structure and is typically and 2. 5/s. The spiral patterns are kept compact (maximum radius
′ ), the scanned area on the sky is only slightly larger than the
about 30′′ /s. The APEX control system currently supports two 2
basic scanning modes: on-the-fly maps (OTF) and spiral scan- FoV and most of the integration time is spent on the central 11′
ning patterns. of the array. These spirals are the preferred observing modes for
pointing scans on sources with flux densities down to a few Jy.
8.1.1. Spirals For fainter sources, the basic spiral pattern can be combined
with a raster mapping mode (raster-spirals) on a grid of pointing
Spirals are done with a constant angular speed and an increasing positions resulting in an even denser sampling of the maps and
radius, therefore the linear scanning velocity is not constant but longer integration time (see Fig. 9, top panel). These compact
increases with time. We have selected two spiral modes of 20 s raster-spirals give excellent results for sources smaller than the

17. 10 G. Siringo et al.: The Large APEX Bolometer Camera LABOCA
FoV of LABOCA and are also suitable for integrations of very
faint sources.
The flexibility in the choice of the spiral parameters also al-
lows spiral observing patterns to be used to map fields much
larger than the FoV. The bottom panel of Fig. 9 shows an exam-
ple of raster of spirals optimized to give an homogeneous cover-
age across a field of 0.5 × 0.5 degrees. In this case, the spirals
start with a large radius and follows an almost circular scan-
ning pattern for each raster position. This mapping mode is very
useful for cosmological deep field surveys since co-adding sev-
eral such raster-spiral scans, taken at different times and thus at
different orientations, provides an optimal compromise between
telescope overheads, uniform coverage and cross-linking of in-
dividual map positions (see Kovács 2008).
8.1.2. On-the-fly maps (OTF)
OTF scans are rectangular scanning patterns produced mov-
ing back-and-forth along alternating rows with a linear constant
speed and accelerating only at the turnarounds. They can be per-
formed in horizontal or equatorial coordinates and the scanning
direction can be rotated relatively to the base system for both co-
ordinate systems. OTF patterns have been tested for maps on the
scales of the FoV up to long slews across the plane of the Milky Fig. 10. Footprint of LABOCA on sky, measured with a beam
Way (2 degrees). Small cross-linked OTFs (of size ∼FoV of map on the planet Mars. The ellipses represent the FWHM shape
LABOCA) give results comparable to the raster-spirals (Kovács of each beam on sky, as given by a two-dimensional Gaussian fit
2008), but the overheads are much larger at a scanning speed of to the single-channel map of each bolometer. Only bolometers
2′ /s. For larger OTFs the relative overheads decrease. with useful signal-to-noise ratios are shown in this map. See also
Fig. 12.
8.2. Ancillary modes
ture, followed by a continuous tip scan in elevation; see also
8.2.1. Point
Sect. 9.5.
The standard pointing procedure consists of one subscan in spi-
ral observing mode and results in a fully sampled map of the FoV
of LABOCA. The pointing offsets relative to the pointing model 9. Performance on the sky and sensitivity
are computed via a two-dimensional Gaussian fit to the source 9.1. Number of channels
position in the map using a BoA pipeline script (see Sect. 10).
Note that this pointing procedure is not limited to pointing scans At the time of the commissioning, the number of channels with
of the central channel of the array but works independently of sky response was 266 (90% of the nominal 295 bolometers). Of
the reference bolometer, thus allowing pointing scans centered these, 6 channels show cross-talk and 12 channels have low sen-
on the most sensitive part of the array. sitivity (less than 10% of the mean responsivity). Two additional
bolometers have been blinded by blocking their horn antennas
with absorber material so they can be used to monitor the tem-
8.2.2. Focus perature fluctuations of the array. The number of channels used
for astronomical observations is therefore 248 (84% yield, see
The default focusing procedure is made of 10 subscans at 5 dif-
Fig. 10).
ferent subreflector positions and 5 seconds of integration time
each. This is the only observing mode without scanning tele-
scope motion. As a result, we are currently restricted to sources 9.2. Array parameters
brighter than the atmospheric variations (Mars, Venus, Saturn
and Jupiter). However, initial tests confirmed that using the wob- Position on sky and relative gain of each bolometer are derived
bler to modulate the source signal allows focusing on weaker from fully sampled maps (hereafter called beam maps) of the
sources, too. This is the only observing mode, so far, for which planets Mars and Saturn (see Fig. 10), besides giving a realistic
the use of the wobbler with LABOCA has been tested. picture of the optical distortions over the FoV. Variations among
maps were found to be within a few arc seconds for the posi-
tions and below 10% for peak flux densities. A table with aver-
8.2.3. Skydips age receiver parameters (RCP)25 is periodically computed from
beam maps and implemented in the BoA software (see Sect. 10).
The attenuation of the astronomical signals due to the atmo- The accuracy of the relative bolometer positions from this mas-
spheric opacity is determined with skydips. These scans mea- ter RCP is typically below 1′′ (5% of the beam size) and the gain
sure the power of the atmospheric emission as a function of the accuracy is better than 5%, confirming the good quality (small
airmass while tipping the telescope from high to low elevation.
A skydip procedure consists of two steps: a hot-sky calibration 25 The latest RCP table is available at:
scan, to provide an absolute measurement of the sky tempera- http://www.apex-telescope.org/bolometer/laboca/calibration/LABOCA-centred

18. G. Siringo et al.: The Large APEX Bolometer Camera LABOCA 11
9.5. Sky opacity determination
Atmospheric absorption in the passband of LABOCA attenuates
the astronomical signals as exp(−τlos ) where the line of sight
optical depth τlos can be as high as 1 for observations at low el-
evation with 2 mm of precipitable water vapor (PWV), typical
limit for observations with LABOCA. The accuracy of the abso-
lute calibration, therefore, strongly depends on the precision in
the determination of τlos .
We use two independent methods for determining τlos . The
first one relies on the PWV level measured every minute by the
APEX radiometer broadly along the line of sight. The PWV
is converted into τlos using an atmospheric transmission model
(ATM, Pardo et al. 2001) and the passband of LABOCA (see
Sect. 4.1). The accuracy of this approach is limited by the knowl-
edge of the passband, the applicability of the ATM and the accu-
racy of the radiometer.
The second method uses skydips (see Sect. 8.2.3). As the
telescope moves from high to low elevation, τlos increases with
airmass. The increasing atmospheric load produces an increasing
total power signal converted to effective sky temperature (T eff )
by direct comparison with a reference hot load. The dependence
Fig. 11. Radial profile of the LABOCA beam derived by av- of T eff on elevation is then fitted to determine the zenith opac-
eraging the beams of all 248 functional bolometers from fully ity τ, used as parameter (see, e.g., Chapman et al. 2004).
sampled maps on Mars. The error bars show the standard devi- Since LABOCA is installed in the Cassegrain cabin of
ation. The main beam is well described by a Gaussian with a APEX, when performing skydips the receiver suffers a wide,
′′ ′′
FWHM of 19. 2 ± 0. 7 and starts to deviate at −20 dB. continuous rotation (about 70 degrees in 20 seconds), which af-
fects the stability of the sorption cooler, thus inducing small vari-
ations of the bolometers temperature (∼1–2 mK). These temper-
ature fluctuations mimic an additional total power signal with
distortion over the entire FoV) of the tertiary optics (see also amplitude comparable to the atmospheric signal. The bolome-
Sect. 2). ters temperature, however, is monitored with high accuracy by
the 3 He-stage thermometer (see Sect. 3.3) and by the two blind
bolometers (see Sect. 9.1), making possible a correction of the
9.3. Beam shape
skydip data.
The LABOCA beam shape was derived for individual bolome- The values of τ resulting from the skydip analysis are robust,
ters from fully sampled maps on Mars (see Fig. 10) as well as yet up to ∼30% higher than those obtained from the radiome-
on pointing scans on Uranus and Neptune. Both methods lead ter. There are several possible explanation for the discrepancy: it
to comparable results and give an almost circular Gaussian with could be the result of some incorrect assumptions going into the
′′ ′′
a FWHM of 19. 2 ± 0. 7 (after deconvolution from the source skydip model (e.g. the sky temperature), or the model itself may
and pixel sizes). We also investigated the error beam pattern of be incomplete. The non-linearity of the bolometers can be an-
LABOCA on beam maps on Mars and Saturn (see Fig. 11). The other factor. The detector responsivities are expected to change
beam starts to deviate from a Gaussian at −20 dB (1% of the peak with the optical load as ∼ exp(−γ τlos ) (to first order in τ), where
intensity). The first error beam pattern can be approximated by a γ can be related to the bolometer constants (Mather 1984) and
Gaussian with a peak of −18.3 dB and a FWHM of 70′′ ± 5′′ , the the optical configuration. Combined with the sky response, the
support legs of the subreflector are visible at the −25 dB (0.3% bolometer non-linearities would increase the effective skydip τ
level). The fraction of the power in the first error beam is ∼18%. by a factor (1 + γ).
Our practical approach to reconciling the results obtained
with the two methods has been to use a linear combination of
9.4. Calibration radiometer and skydip values such that it gives the most consis-
tent calibrator fluxes at all elevations27. The excellent calibra-
The astronomical calibration was achieved on Mars, Neptune tion accuracy of LABOCA (see Appendix A) underscores this
and Uranus and a constant conversion factor of 6.3 ± approach.
0.5 Jy beam−1 µV−1 was determined between LABOCA re-
sponse and flux density. For the determination of the calibra-
tion factor we have used the fluxes of planets determined with 9.6. Sensitivity
the software ASTRO26 . The overall calibration accuracy for
LABOCA is about 10%. We have also defined a list of sec- The noise-weighted mean point-source sensitivity of the array
ondary calibrators and calibrated them against the planets (see (noise-equivalent flux density, NEFD) determined from on-sky
Appendix A). In order to improve the absolute flux determina- integrations, is 55 mJy s1/2 (sensitivity per channel). This value
tion, the calibrators are observed routinely every ∼2 hours be- is achieved only by filtering the low frequencies (hence large
tween observations of scientific targets.
27 A text file (BoA readable, see Sect. 10) containing the zenith
opacities calculated from the skydips and the radiometer is available at:
26 GILDAS package, http://www.iram.fr/IRAMFR/GILDAS http://www.apex-telescope.org/bolometer/laboca/calibration/opacity/

19. 12 G. Siringo et al.: The Large APEX Bolometer Camera LABOCA
scale emission) to reject residual sky-noise. For extended emis-
sion, without low frequency filtering, there is a degradation of
sensitivity to a mean array sensitivity of 80–100 mJy s1/2 de-
pending on sky stability. However, there are significant varia-
tions of the sensitivity across the array (see Fig. 12, top).
For detection experiments of compact sources with known
position, LABOCA can be centered on the most sensitive part of
the array rather than on the geometric center. This results in an
improved point source sensitivity of ∼40 mJy s1/2 for compact
mapping pattern like spirals.
9.7. Mapping speed and time estimate
The relation between the expected residual rms map noise σ,
the surveyed area on the sky and the integration time can be ex-
pressed as
2
(Xscan + D) (Yscan + D) fsamp NEFD eτlos
tint = (1)
Abeam Nbol σ
where Xscan , Yscan are the dimensions of the area covered by the
scanning pattern, D is the size of the array, Abeam is the area of
the LABOCA beam, Nbol is the number of working bolometers,
tint is the on source integration time, NEFD is the average array
sensitivity, fsamp is the number of grid points per beam and τlos 50
the line-of-sight opacity. This formula28 does not include sensi-
tivity variations across the array and the increasing sparseness
of data points toward the map edges (the latter effect depends
on the mapping mode). We have tested this estimate on deep 40
integrations for point like and extended sources and found a rea-
sonable agreement in the measured rms noise of the processed
maps.
30
Pixel Count
9.8. Noise behavior in deep integrations
In order to study how the noise averages down with increasing 20
integration time, about 350 hours of blank-field mapping obser-
vations have been co-added with randomly inverted signs to re-
move any sources while keeping the noise structure. As shown
in Fig. 13, the noise integrates down with t−1/2 , as expected. 10
10. Data reduction
0
LABOCA data are stored in MB-FITS format (Multi-Beam
FITS) by the data writer embedded in APECS. A new software 20 40 60 80 100 120 140
package has been specifically developed to reduce LABOCA 1/2
Point Source NEFD (mJy s )
data: the Bolometer array data Analysis software (BoA). It is
mostly written in the Python language, except for the most com- Fig. 12. Top: Effective point source sensitivities (after low fre-
puting demanding tasks, which are written in Fortran90. quencies filtering) of all the useful bolometers of LABOCA
BoA was first installed and integrated in APECS in early (color scale in Jy beam−1 ). The positions were determined from
2006. An extensive description of its functionalities will be given beam maps on Mars, thus providing a realistic picture of the
in a separate paper (Schuller et al. in prep.). In this Section, optical distortions over the FoV (see also Fig. 10). Bottom:
we outline the most important features for processing LABOCA Distribution of the effective point source sensitivities on the
data. LABOCA array (5 mJy s1/2 binning). The total mapping speed
of the array is as if the 250 good bolometers were all identical
at a level of 54.5 mJy beam−1 s1/2 (thick black line). The median
10.1. On-line data reduction sensitivity is also shown (dotted line).
During the observations carried out at APEX, the on-line data
calibrator (as part of APECS) performs a quick data reduction
for the basic pointing and focus observing modes, for which the
of each scan to provide the observer with a quick preview of the
on-line calibrator computes and sends back to the observer the
maps or spectra being observed. This is of particular importance
pointing offsets or focus corrections to be applied. For both fo-
28 An online time estimator is available at: cus and pointing scans, only a quick estimate of the correlated
http://www.apex-telescope.org/bolometer/laboca/obscalc/ noise (see below) is computed and subtracted from the data. The

20. G. Siringo et al.: The Large APEX Bolometer Camera LABOCA 13
Fig. 13. Noise behavior for deep integrations has been studied
by co-adding about 350 hours of blank-field data. The plot shows
the effective residual rms noise on sky.
focus correction is derived from a parabolic fit to the peak flux
measured by the reference bolometer as a function of the sub-
reflector position. For pointing scans, the signals of all usable
channels are combined into a map of the central 5′ × 5′ area, in
horizontal coordinates. A two-dimensional elliptical Gaussian is
then fitted to the source in this map, which gives the pointing
offsets, as well as the peak flux and the dimensions of minor and Fig. 14. Time series of a bolometer signal on blank sky at con-
major axis of the source. secutive steps during the data reduction process with BoA. The
signals, labeled from 1 to 5, are shown for: 1) original signal on
sky; 2) after correcting for system temperature variations using
10.2. Off-line data reduction the blind bolometers (∼200 µK during this scan); 3) after median
skynoise removal over the full array, one single iteration; 4) after
The BoA software can also be used off-line to process any kind
median skynoise removal over the full array, 10 iterations; 5) af-
of bolometer data acquired at APEX. The off-line BoA runs in
the interactive environment of the Python language. In a typ- ter correlated noise removal, grouping the channels by amplifier
and cable, 5 iterations (shifted to a −3 level, for clarity).
ical off-line data reduction session, several scans can be com-
bined together, for instance to improve the noise level on deep
integrations, or to do mosaicing of maps covering adjacent ar-
eas. The result of any data reduction can be stored in a FITS means that the signal that they recorded is not used any fur-
file using standard world coordinate system (WCS) keywords, ther in the processing.
which can then be read by other softwares for further processing – Flagging of stationary points. The data acquired when the
(e.g. source extraction, or overlay with ancillary data). telescope was too slow to produce a signal inside the use-
The common steps involved in the processing of LABOCA ful part of the post-detection frequency band of LABOCA
data are the following: (e.g. below 0.1 Hz, see Sec. 8.1 and Fig. 8) can be flagged.
Data obtained when the telescope acceleration is very high
– Flux calibration. A correct scaling of the flux involves, at may show some excess noise, and can also be flagged.
least, two steps: the opacity correction and the counts-to-Jy – Correlated noise removal. This can be done using a Principal
conversion. The zenith opacity is derived from skydip mea- Components Analysis (PCA), or a median noise removal
surements (see Sect. 8.2.3 and Sect. 9.5), and the line of sight method. In the latter case, the median value of all (nor-
opacity also depends on the elevation. The counts-to-Jy fac- malized) signals is computed at each timestamp, and sub-
tor has been determined during the commissioning of the in- tracted from the signal of each channel (with appropriate
strument, but additional correction factors may be applied, relative gains). This can be performed using all beams at
depending on the flux measured on calibrators with known once or better on groups of selected beams. In fact, some
fluxes (see Sect. 9.4). groups of channels sharing the same electronics subsystem
– Flagging of bad channels. Bolometers not responding or with (e.g. amplifier box, flat cable) can show strong correlation
strong excess noise are automatically identified from their and removing the median signal on those groups of chan-
rms noise being well outside the main distribution of the rms nels greatly improves their signal-to-noise ratio. However,
noise values across the array. They can be flagged, which it should be noted that an astronomical source with extended

21. 14 G. Siringo et al.: The Large APEX Bolometer Camera LABOCA
uniform emission on scales of several arcmin or larger would
mimic the correlated noise behavior of groups of channels.
Therefore, subtracting the median noise results in filtering
out some fraction of the extended emission, depending on
the size and morphology of the source.
– Despiking. Data points that deviate by more than a given
factor times the standard deviation in each channel can be
flagged.
– Data weighting and map making. To build a map in hori-
zontal or equatorial coordinates, the data of all usable chan-
nels are projected onto a regular grid and co-added, using a
weighted average, with weights computed as 1/rms2 , where
c
rmsc refers to the rms noise of an individual channel. The
channel rms noise can either be the standard deviation of the
data over the full time line, or it can be computed on a sliding
window containing a given number of integrations.
A visual example of correlated noise removal with BoA is given
in Fig. 14.
Additional (optional) steps that can improve the final reduc-
tion include: removing slow variations from the signal, by sub-
tracting a polynomial baseline or by filtering out low frequencies
in the Fourier domain; smoothing of the map with a Gaussian
kernel of a given size.
The map resulting from a full reduction can be used as a
model of the astronomical source to mask the data, in order to
repeat some computations without using data points correspond-
ing to the source in the map. This iterative scheme helps in the
difficult task to recover the extended emission and reduces the
generation of artifacts around strong sources. In the presence of
bright sources in the map, a typical data reduction should first
perform the reduction steps as listed above with conservative
parameters (for example, using high enough thresholds in the
despiking to avoid flagging out the strong sources), and should
then read again the raw data, use the resulting map to temporarily
flag out datapoints corresponding to bright sources, and repeat
the full process with less conservative parameters.
In addition to the data processing itself, BoA allows visual-
ization of the data in different ways (signal vs. time, correlation
between channels, power spectra or datagrams), as well as the
telescope pattern, speed and acceleration. Finally, BoA also in-
cludes a simulation module, which can be used to investigate
the mapping coverage for on-the-fly maps, spirals and raster of Fig. 15. A map of the galactic center, extracted from the project
spirals, given the bolometer array parameters. ATLASGAL (F. Schuller et al., in prep.). The map is 2×4 de-
grees large, the residual noise is 50 mJy beam −1 and it required
only 6 hours of observing time. Data reduced with the BoA pack-
age.
11. Science with LABOCA
Because of its spectral passband, centered at the wavelength of 11.2. Star Formation in the Milky Way
870 µm (see Fig. 5), LABOCA is particularly sensitive to ther-
mal emission from cold objects in the Universe which is of great The outstanding power of LABOCA in mapping large areas of
interest for a number of astrophysical research fields. sky with high sensitivity (see Fig. 15) makes it possible, for the
first time, to perform unbiased surveys of the distribution of the
cold dust in the Milky Way. As the dust emission at 870 µm is
11.1. Planet Formation typically optically thin, it is a direct tracer of the gas column
density and gas mass. Large scale surveys in the Milky Way
The study of Kuiper Belt Objects in the Solar System as well will reveal the distribution and gas properties of a large num-
as observations of debris disks of cold dust around nearby main ber of pre-star cluster clumps and pre-stellar cores in different
sequence stars can give vital clues to the formation of our own environments and evolutionary states. Equally importantly, they
planetary system and planets in general. With its angular resolu- provide information on the structure of the interstellar medium
tion of 19′′ (see Fig. 11) LABOCA can resolve the debris disks on large scales at high spatial resolution, little explored so far.
of some nearby stars. Such surveys are vital to improve our understanding of the pro-

22. G. Siringo et al.: The Large APEX Bolometer Camera LABOCA 15
Fig. 16. Map of the nearby galaxy Centaurus A (NGC 5128,
from Weiß et al. 2008). The central dust disk has a pronounced Fig. 17. LABOCA images of SMM J14011+0252 (left)
S-shape. Our LABOCA image shows, for the first time at sub- and SMM J14009+0252 (right) smoothed to 25′′ resolution.
millimeter wavelengths, the synchrotron emission from the radio SMM J14011 is at a redshift of z = 2.56 while SMM J14009 has
jets and the inner radio lobes. Its distribution follows closely the no reliable redshift determination. The noise level of the map is
known radio emission at cm wavelengths. The color scale is in about 2.5 mJy beam−1 .
units of Jy beam−1 . The residual rms noise is 4 mJy beam−1 and
the total integration time is 5 hours.
and thus, in contrast to the local Universe, luminous obscured
galaxies at high redshift could dominate the total bolometric
cesses that govern star formation as well as the relation between emission from all galaxies at early epochs. The mass-tracing
the clump mass spectrum and the stellar initial mass function. property of submillimeter dust emission (see Sect. 1.1) allows
Large unbiased surveys are also critical for finding precursors us to make a direct estimate of the star formation rates (SFRs) of
of high-mass stars, which are undetectable at other wavelengths these objects. The generally high observed SFRs suggest that ap-
due to the high obscuration of the massive cores they are embed- proximately half of all the stars that have formed by the present
ded in. LABOCA will help to obtain a detailed understanding of day may have formed in highly obscured systems which remain
their evolution. In addition, deep surveys of nearby, star-forming undetected at optical or near-infrared wavelengths. One example
clouds, will allow study of the pre-stellar mass function down to for such system is the submillimeter galaxy SMM J14009+0252
the brown dwarf regime. (see Fig. 17), which is strong in the submillimeter range, has a
1.4 GHz radio counterpart, but no obvious counterpart in deep K-
band images (Ivison et al. 2000). Clearly it is critical to include
11.3. Cold gas in Galaxies these highly obscured sources in models of galaxy formation to
The only reliable way to trace the bulk of dust in galaxies is obtain a complete understanding of the evolution of galaxies.
through imaging at submillimeter wavelengths. It is becoming With its fast mapping capabilities LABOCA allows us to map
clear that most of the dust mass in spiral galaxies lies in cold, fields of a half square degree, a typical size of a deep cosmolog-
low surface-brightness disks, often extending far from the galac- ical field surveyed at other wavelengths, down to the confusion
tic nucleus, as in the case of the starburst galaxy NGC 253 limit in a reasonable amount of observing time. This will also
or of NGC 5128 (Cen A, see Fig. 16 and Weiß et al. 2008). greatly improve the statistics of high redshift galaxies detected
Understanding this component is critically important as it dom- at submillimeter wavelengths. See Beelen et al. (2008) for the
inates the total gas mass in galaxies and studies of the Schmidt first deep LABOCA cosmology imaging.
law based on H i observations only can heavily underestimate
the gas surface density in the outer parts.
In addition to studying individual nearby galaxies, LABOCA 12. Future plans
will be vital for determining the low-z benchmarks, such as the In collaboration with the Institute for Photonics Technology
local luminosity and dust mass functions, which are required to (IPHT, Jena, Germany) we are working on a new bolometer ar-
interpret information from deep cosmological surveys. ray, LABOCA-2, using superconducting technology. The new
array will use superconducting thermistors (transition edge sen-
11.4. Galaxy formation at high redshift sor, TES), planar dipole absorbers, SQUID (Superconducting
Quantum Interference Device) multiplexing and amplifica-
Due to the negative-K correction, submillimeter observations of- tion. A system already using the same technology, called
fer equal sensitivity to dusty star-forming galaxies over a redshift Submillimeter APEX Bolometer Camera (SABOCA) is going to
range from z ∼1–10 and therefore provide information on the star be commissioned as facility instrument on the APEX telescope
formation history at epochs from about half to only 5% of the for operation in the 350 µm atmospheric window. Additionally,
present age of the universe. Recent studies have shown that the given the low sensitivity of superconducting bolometers to mi-
volume density of luminous submillimeter galaxies (SMGs) in- crophonics, it will be possible to move from the wet cryostat to
creases over a thousand-fold out to z ∼2 (Chapman et al. 2005), a pulse-tube cooler, as already successfully tested at MPIfR in